Steering Through the Inner Residue

Multi-channel video installation, 6 LED Panels, Digital Prints

2026

What happens when a system confronts its own output?

By recursively feeding generated images back into generative image models, Steering through the Inner Residue makes visible the characteristics and nature of image-making processes in artificial intelligence.

Artificial intelligence is often portrayed as a black box with clearly defined inputs and outputs. Most image models are initialized with a textual description (prompt) that guides their outcomes but also reflects human expectations. By disregarding the text entirely and guiding it only by its own image output, the underlying algorithmic processes become visible.

The six image streams resemble an unfolded box, each surface carrying reverberations of the process – a landscape of distinctive structures emerging from the procedures. What accumulates over time is not an image but a process: an algorithmic pattern, a contingent residue that signifies nowhere else but the process itself. Yet through human reception of their generated outputs, these processes increasingly feed back into the very world they are meant to mimic.

A collaboration between Leon-Etienne Kühr & Ting-Chun Liu